Projet Image10k - Images10k - an open, large, human-annotated dataset for biological and artificial visual processing

Images10k is an open, human-annotated dataset comprising 8,382 high-quality, natural images across 15 object categories. Each image portrays its subject in its typical environment and has been validated via online behavioral experiments to ensure label accuracy and contextual relevance.

On this page, you’ll find:

Interactive Carousels – Image galleries organized by semantic category for easy browsing.
Image Collection & Annotation – A step-by-step overview of our data acquisition and labeling pipeline.
Summary Statistics & Validation – Key metrics and behavioral results confirming dataset reliability.
Metadata Showcase – Detailed metadata for each image, including labels, category assignments, and validation scores.

Images10k - an open, large, human-annotated dataset for biological and artificial visual processing

Introduction